Flemish 4G models

De 4th generation transport model for passenger transport in Flanders is entirely agent-based, in design, as well as estimation and application.


The strategic transport model is used as the leading transport model which is used to distract derivative models. And is used as a testing ground for the development of structural extensions and functionalities of the model. In the 4G project a total re estimation of the model was executed, with Visum-LOS as starting point. Essential extensions were finalized: an extensive travel time and activity duration model, variable use of time periods within the models, fragmented park-and-ride modelling, expansion of person characteristics, tuning of sub processes and spatial differences between regions. The total model is efficiently coded in the .NET-system. In addition the spm Flanders v4.1 was used for the first time for the support of the MPV2030, where functionalities on road pricing were evaluated. All these experiences and developments were used to hold a thorough training session for internal and external colleagues.

From version 4.2 onward the model software Visum is used for the maintenance of the networks, including network assignment and the derivation of LOS for motorized vehicles and public transport.  Despite the ambitious size of the strategic models, semi-dynamic assignment is chosen as method for motorized vehicles, using blocking back and spill-over of waiting queues over adjacent periods. For PT is opted to include roof topping, to integrate the possibilities of timetable based Visum in the agent-based modelling approach. The functionalities of Visum are extended by personalized developed applications in the .NET-system. Calibration is performed on agent-level and exceeds classical matrix calibration: simultaneously over modes and periods the exact tours are calibrated by using an optimizer of our own. The results of this calibration step are used for structural adjustments in parts of the demand model.

The 4G modelling system is a starting point for the implementation of a centralized coordination meta-dataset, which forms a base for a family of derivative transport models. This guarantees consistency between the model instruments and ensures the user with quality and up-to-date data. An efficient process is developed to shape derivative transport models, with distinction between leading and derivative models.  This process can automatically create derivative zoning, supply networks for motorized vehicles and PT, socioeconomic data, and starting patterns will be tailor-made. The model can be applied by use of the user interface MIG4, which guides the user through the creation and maintenance of scenarios, the evaluation of effects and the analysis of the results using Visum and the RSL-Viewer. Beneath the surface the RekenProcessor (calculation processor) maintains all the model processes and designs scripts that will model specifically what the user needs.